A Novel Functional Interaction between the Sp1-like Protein KLF13 and SREBP-Sp1 Activation Complex Underlies Regulation of Low Density Lipoprotein Receptor Promoter Function*

Cholesterol homeostasis is regulated by a family of transcription factors designated sterol regulatory element-binding proteins (SREBPs). Precise control of SREBP-targeted genes requires additional interactions with co-regulatory transcription factors. In the case of the low density lipoprotein receptor (LDLR), SREBP cooperates with the specificity protein Sp1 to activate the promoter. In this report, we describe a novel pathway in LDLR transcriptional regulation distinct from the SREBP-Sp1 activation complex involving the Sp1-like protein Krueppel-like factor 13 (KLF13). Using a combination of RNA interference, electrophoretic mobility shift, chromatin immunoprecipitation, and reporter assays, deletion, and site-directed mutagenesis, we demonstrated that KLF13 mediates repression in a DNA context-selective manner. KLF13 repression of LDLR promoter activity appears to be needed to keep the receptor silent, a state that can be antagonized by Sp1, SREBP, and inhibitors of histone deacetylase activity. Chromatin immunoprecipitation assay confirmed that KLF13 binds proximal LDLR DNA sequences in vivo and that exogenous oxysterol up-regulates such binding. Together these studies identify a novel regulatory pathway in which gene repression by KLF13 must be overcome by the Sp1-SREBP complex to activate the LDLR promoter. Therefore, these data should replace a pre-existent and more simple paradigm that takes into consideration only the induction of the activator proteins Sp1-SREBP as necessary for LDLR promoter drive without including default repression, such as that by KLF13, of the LDLR gene.

Granulosa (Ovarian) Cell Culture-Ovaries from pre-pubertal (60 -70 kg) swine were collected at an abattoir and transported to the laboratory in iced saline. Granulosa cells were isolated from small-and medium-sized (1-5 mm) antral Graafian follicles by fine needle aspiration under sterile conditions and washed three times by low speed centrifugation (3000 revolutions/min) in MEM. Approximately 5 ϫ 10 6 viable cells were plated in 12-well culture dishes (Corning, NY) containing bicarbonate-buffered MEM and 3% fetal bovine serum with insulin (1 g/ml), estradiol(0.5 g/ml), and follicle-stimulating hormone (5 ng/ml) to permit anchorage and partial luteinization (20). Cells were allowed to attach to culture dishes for 48 h at 37°C in 5% CO 2 .
Transient Transfection-In transient transfection analyses, we utilized a 1087-bp 5Ј upstream regulatory fragment (Ϫ1076 to ϩ11 bp from the transcriptional start site) of the porcine LDLR and 5Ј-nested deletional constructs driving a cytoplasmically targeted firefly luciferase cDNA (20,21,29). After attachment of granulosa cells for 48 h (see above), hormone-free medium was replaced every 24 h twice. Before transfection, monolayers were incubated in serum-free MEM without antibiotics for 20 -30 min. Transfection medium (1 ml/well) comprised serum-free MEM without antibiotics containing 1 g of total plasmid DNA and 6 l of Lipofectamine. Based on prior time course experiments, 5% serum-containing medium was replaced after 6 h of transfection. After an additional 24 h of recovery to allow vector expression, cells were exposed to serum-free MEM containing antibiotics and the indicated inhibitors or vehicle for 4 or 24 h to monitor basal activity and responses to trichostatin A (TSA) or sodium butyrate, respectively. Where indicated, cells were transfected with pcDNA3.1/HisC epitopetagged KLF13, pcDNA3.1/SREBP-1a (sterol-responsive element-binding protein), and/or pCMV/hemagglutinin epitope-tagged p300. To quantify reporter expression, cultures were rinsed once at room temperature with Dulbecco's phosphate-buffered saline, lysed in 100 l of 1ϫ lysis buffer (Luciferase Assay System, Promega, Madison, WI), and stored at Ϫ70°C until later assay. Luciferase activity was measured using 100 l of firefly substrate (Promega) and 20 l of cellular lysate in a Turner TD-20/20 luminometer (Turner Designs, Sunnyvale, CA). Data are normalized as relative light units/100 g of protein and expressed as the mean Ϯ S.E. All experiments were performed at least three times with triplicate incubations.
Transfection of mutant LDLR promoter sequences was performed with equimolar concentrations of DNA. A promoterless luciferase construct, p0/luc, exhibiting no significant activity in response to any intervention, was used to adjust total DNA to 1 g. In co-transfection studies, 0.6 g of full-length LDLR promoter (pLDLR1076/luc) was added with 0.01-0.3 g of pcDNA3.1/KLF13 (10), pCMV empty vector and/or pCMV/SREBP-1a (28) (obtained from Dr. Timothy F. Osborne, Department of Molecular Biology and Biochemistry, University of California, Irvine, CA), or pCMV/Sp1 or Sp3 (from Dr. Robert Tjian, University of California, Berkeley).
Nuclear Protein Isolation and Electrophoretic Mobility Shift Assay (EMSA)-The general procedure outlined by Dignam et al. (50) was followed with some modifications, as follows. Granulosa-luteal cells were washed twice with cold phosphate-buffered saline and detached from culture dishes by scraping, recovered by centrifugation at 500 ϫ g for 5 min at 4°C, resuspended in ice-cold phosphate-buffered saline, pH 7.4, and pelleted at 12,000 ϫ g for 20 s. The cells were lysed in ice-cold buffer A (10 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, and freshly added 10 l of protease inhibitor mixture (Sigma)) and mixed by pipetting. After incubation on ice for 10 min, Nonidet P-40 was added to a final detergent concentration of 0.05% and the solution mixed by pipetting before centrifugation at 12,000 ϫ g for 20 s. The nuclear pellet was suspended in 100 l of ice-cold buffer B (containing 20 mM HEPES-KOH, pH 7.9, 1.5 mM MgCl 2 , 0.4 M KCl, 0.2 mM EDTA, 0.5 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, 25% glycerol, and 5 l of protease inhibitor mixture). The suspension was placed on ice for 30 min with an occasional gentle shaking and then centrifuged at 12,000 ϫ g for 15 min at 4°C to obtain the nuclear extract (supernatant). The latter was stored at Ϫ70°C. Protein concentrations were measured by the Bradford method (Bio-Rad).
EMSA reactions were performed in 20 l of 10 mM Tris-HCl, pH 7.5, 50 mM NaCl, 1 mM EDTA, 5% glycerol, 50 g/ml poly(dI-dC), 32 Plabeled oligonucleotide probe (50,000 counts/min), and 10 g of nuclear protein for 30 min at 23°C. Antibodies or cold oligonucleotides were pre-incubated with nuclear proteins at 4°C for 1 h or for 30 min, respectively. Protein-DNA complexes were separated from free probe by 5% non-denaturing polyacrylamide gel electrophoresis at 200 V for 2.5 h. The gels were subjected to autoradiography overnight at Ϫ80°C.
Chromatin Immunoprecipitation (ChIP) Assay-ChIP assay evaluates whether a given transcription factor is bound to a distinct proximal promoter DNA sequence in living cells. Specificity of transcription factor binding is achieved by immunoprecipitation of formaldehyde-crosslinked protein-DNA (chromatin) complexes. The gene sequence is verified by primer-specific PCR of DNA purified from chromatin precipitates. For ChIP assay, granulosa-luteal cells were transfected with epitope-tagged KLF13 or p300 and then consecutively cross-linked by the addition of formaldehyde at a 1% final volume for 10 min at 37°C, quenched by adding glycine, and harvested by scraping in ice-cold phosphate-buffered saline with fresh complete protease inhibitors (Roche Applied Science). The cells were then pelleted and resuspended in SDS lysis buffer containing inhibitors and then sonicated on ice to release 200 -800-bp DNA fragments, microcentrifuged for 10 min at 4°C, and resuspended in dilution buffer. The cells were pre-cleared with DNAand albumin-blocked protein A-agarose for 30 min at 4°C, aliquoted to retain a 5% sample for later PCR of DNA input, and incubated overnight with affinity-purified Ab to KLF13 or cyclic AMP-response elementbinding protein versus no Ab versus mouse gamma globulin (negative controls) with rotation at 4°C and then precipitated with salmon sperm DNA protein A-agarose for 1 h at 4°C. The cells were washed successively in low and high salt, LiCl and Tris-EDTA buffers (Cell Signaling Technologies catalog number 17-295, Lake Placid, NY). Cross-linked DNA complexes were dissociated for real-time-PCR (see below) or boiled for 10 min in Laemmli buffer for Western blot.
To reverse cross-links, chromatin complexes (including input samples) were incubated in 5 M NaCl at 65°C for 4 h, resuspended in Tris-EDTA-proteinase K buffer for 1 h at 45°C, extracted in triple alcohol, and reprecipitated with 20 g of yeast tRNA and 2 volumes of ethanol overnight at 4°C. The complexes were then washed with 70% ethanol and air dried. Purified DNA was resuspended in 25 l of H 2 O for PCR using the LDLR promoter-specific primers (ϩ), 5Ј-GAGTCAGGGCTTC-ACGGGTTA-3Ј and (Ϫ), 5Ј-CTGTTCACTGTGTGCGCTCTTG-3Ј.
Western Blot-Western blots were performed by boiling nuclear protein extracts in Laemmli buffer, separation on 10% denaturing SDS-PAGE, transfer to nitrocellulose membranes, and immunochemiluminescent detection as described previously (29).
Small Interfering RNA (siRNA)-KLF13 RNA interference was achieved using a pool of four siRNA duplexes obtained from Dharmacon Research (SMARTpool, human KLF13, MN_015995, Lafayette, CO). Non-targeting siRNA served as a control DNA sequence with no matches upon BLAST search. Transfection of granulosa-luteal cells with siRNA was performed using Lipofectamine reagent (Invitrogen) to achieve a final concentration of siRNA of 20 nM (optimized from 10, 20, 40, and 100 nM). Transfection also included empty vector, full-length pLDLR1076/luc, and/or human KLF13 followed by recovery for 48 h before assay of luciferase activity. Analogous transfection of HepG2 cells was used to extract protein for Western blotting to confirm inhibition of KLF13 expression.
Statistical Methods-Observations are based on three or more independent experiments conducted using separate batches of 150 -200 ovaries. Data pooled among experiments were subjected to one-or two-way analysis of variance in a repeated measures design (30). Ratio values (observed to control) were log-transformed to limit the dispersion of residual variance. Means were contrasted by the post hoc Tukey multiple comparison test at p Ͻ 0.05.

RESULTS
The LDLR gene promoter used in this study exhibits three Sp1-like TC-rich (5Ј-TCCTCC-3Ј) sequences flanking a canonical sterol response element (SRE) within a delimited Ϫ255/Ϫ139-bp region 5Ј upstream of the transcriptional start site (21). Each TC-predominant sequence binds immunoreactive Sp1, Sp3, and an unidentified mithramycin-displaceable nuclear protein (29,31). In the current study, initial experiments aimed at characterizing the importance of these sites in the transcriptional control of LDLR promoter revealed sensitivity to HDAC inhibitors, suggesting a role for histone modification. As shown in Fig.  1A, porcine granulosa-luteal cells, used here as a model for cells that both contain the LDLR and display an intense cholesterol metabolism involved in hormone synthesis, were exposed to the HDAC inhibitors TSA (10 ng/ml) or sodium butyrate (0.5 mM) (32) for 24 h after transfection with pLDLR1076/luc or selected proximal deletional constructs. TSA and sodium butyrate increased reporter activity of full-length and truncated pLDLR455/luc and pLDLR255/luc by Ͼ2-fold. Quantifiable real-time-PCR revealed that TSA also increased LDLR transcript abundance by 2.2-fold (not shown). Deletion to pLDLR139/luc abrogated TSA and sodium butyrate-induced up-regulation, pointing to an acetylation-sensitive inhibitory region located Ϫ255/Ϫ139 bp 5Ј upstream of the transcriptional start site. This segment in the pig LDLR gene contains three putative Sp1 elements surrounding a canonical SRE (Fig. 1B). The Sp1-like motif is 5Ј-TCCTCC-3Ј, which is analogous to DNA sequences by which Sp1-KLF proteins activate or repress other genes (3,33). In earlier gel mobility shift assay analyses, each of the immunoreactive Sp1 and Sp3 and an unknown protein extracted from the nuclei of granulosa-luteal cells bound all three Sp1-like sequences in the porcine LDLR promoter (29). Therefore, these results lead us to hypothesize that the unidentified nuclear protein that binds the TC-rich DNA sequences in the proximal LDLR promoter may be a member of the Sp/KLF family of proteins, because they share similar binding sequences.
To test this hypothesis, we first assessed Sp/KLF gene expression in ovarian cells by real-time-PCR using low stringency primers to conserved zinc finger domains of the KLF superfamily. The motivation was recent microarray data showing transcripts for KLF2, KLF4, and KLF9 in the ovary and testis, although of unknown function (34 -36). Quantifiable PCR and sequencing revealed transcripts for KLF9 (GenBank TM AY850383), KLF4 (GenBank TM DQ000310), and KLF13 (GenBank TM AY850382) with relative abundancies of 1.0, 48, and 25, respectively (n ϭ 10 experiments) (not shown). The swine KLF9 and KLF4 coding sequences comprised 735 and 1533 nucleotides and had 96.7 and 89.2% sequence identity with cognate human genes, respectively. The KLF13 cDNA contained 879 nucleotides and had 83% identity with the human sequence. Predicted KLF9 (244 amino acids), KLF4 (512 amino acids), and KLF13 (292 amino acids) peptides were 99.2, 92.4, and 92.8% identical with the human counterparts, respectively (evaluated by EMBOSS-Align www.ebi.uk). These data indicate that these KLF proteins are good candidates to regulate the LDLR promoter in these cells.
Earlier EMSA analyses performed with nuclear extracts of granulosaluteal cells identified competitive binding by immunoreactive Sp1 and Sp3 to each of three contextually distinct LDLR promoter TC-rich oligodeoxynucleotide sequences in the LDLR promoter (29). Our early study confirmed that granulosa-luteal cell nuclear protein binds to all three TC-rich motifs with Sp1/Sp3 immunospecificity. There was an unknown, rapidly migrating complex, which was postulated to reflect a Sp/KLF protein. Here, we show that GST-tagged porcine KLF13 binds each of the above three radiolabeled oligos competitively to form a single protein-DNA complex with maximal binding to the 5Ј-most upstream (Ϫ226/Ϫ202) Sp1-like sequence (Fig. 2). Specificity was confirmed by demonstrating that the pre-addition of antibody to KLF13 or of cold oligos diminishes, whereas deletion of the C-terminal triple zinc finger DNA-binding domain abolishes, formation of EMSA complexes. In three experiments, competition with 100-fold molar excess cold oligo or with KLF13-specific antibody (RFLAT-1) reduced radiographic density (percentage of 100% density in the absence of cold oligo or antibody) to 15 and 16% (for the Ϫ127 to Ϫ155 oligos), to 52 and 4% (for the Ϫ184 to Ϫ207 oligos), and to 18 and 7% (for the Ϫ202 to Ϫ226 oligos). Preaddition of KLF13 antibody disrupted complexes formed between protein and Ϫ127 to Ϫ155 double-stranded DNA without supershifting the same. Thus, these results indicate that KLF13 binds strongly and specifically to Sp1-like cis-DNA sequences present in porcine LDLR.
Transfection of a full-length cDNA encoding pig KLF13 into ovarian granulosa-luteal cells (10, 30, 100 and 300 ng DNA) suppressed basal pLDLR1076/luc activity (Fig. 3A). Co-transfection of KLF13 (300 ng/well) uniformly suppressed basal pLDLR-luc activity by Ͼ75% (absolute range in 12 experiments 78 -95%; median 87%). Human KLF13 acted in the same manner (not shown). Interestingly, the extent of inhibition by KLF13 was cell-selective, because comparative experiments in HepG2 cells (n ϭ 7 experiments) revealed fractional repression of 10 -25%. Exposure to the HDAC inhibitor TSA (32) for 24 h, beginning immediately after KLF13 transfection, overcame reporter repression fully. ChIP assay verified that KLF13 binds to proximal LDLR promoter DNA sequences in vivo in granulosa-luteal cells. In particular, the antibody to either the His epitope tag or to the transfected KLF13 peptide precipitated nuclear chromatin containing PCR-proximal LDLR promoter sequences (Fig. 3B). The specificity of cellular KLF13 as a repressor of the LDLR promoter was tested using (highly) selective siRNAs to silence KLF13 in granulosa-luteal cells. Active and control (unmatched) siRNAs exerted a 1.5-fold stimulatory effect (p Ͻ 0.05) and no effect on basal pLDLR1076/luc activity (n ϭ 4 experiments), respectively (Fig.  3C). Active siRNAs fully reversed suppression by exogenous KLF13 (p Ͻ 0.01 versus control siRNAs). Western blot verified Ͼ95% reduction of cellular KLF13 protein by cognate (but not control) siRNAs. The addition of siRNAs to Sp1 reduced promoter stimulation by Sp1 but did not alter inhibition by KLF13 (n ϭ 3 experiments, data not shown). To examine the possible involvement of KLF13 in LDLR repression by oxysterol, granulosa-luteal cells were incubated with 25-hydroxycholesterol (1 M, a one-half maximal inhibitory concentration). As shown in Fig. 3D, chromatin immunoprecipitation assay disclosed that endogenous KLF13 binds to the LDLR promoter only in the presence of oxysterol. Thus, the results of ChIP assays and siRNA-mediated knockdown clearly indicate that KLF13 binds and represses the LDLR promoter in vivo.
A total of three Sp1-like TC-rich cis-DNA motifs in the pig LDLR promoter reside 5Ј upstream (two motifs) and 3Ј downstream (one motif) of an intervening SRE (Fig. 1B). To test which of these sites is FIGURE 2. EMSA using the GST-KLF13 fusion protein and the three indicated radiolabeled double-stranded oligodeoxynucleotides containing TC-rich cis-DNA sequences within the proximal swine LDLR promoter. Oligo probes were incubated with or without recombinant human GST-KLF13, competing cold oligos, or antibody (Ab) to KLF13. C-terminally deleted KLF13 peptides (comprising residual amino acids 1-172 (1) and amino acids 1-35 (2)) were added in the two right lanes. Gels are representative of three comparable experiments. FEBRUARY 10, 2006 • VOLUME 281 • NUMBER 6 more sensitive to KLF13 and whether there is a context-specific effect, we introduced bp transversions in each Sp1-like motif as well as the intervening SRE, yielding pLDLR148mut/luc, Ϫ198mut/luc, Ϫ213mut/ luc, and Ϫ161SREmut/luc. Compared with wild-type pLDLR1076/luc, basal expression of pLDLR148mut/luc and pLDLR198mut/luc was reduced by Ͼ85% (p Ͻ 0.01) and Ͼ50% (p Ͻ 0.05), respectively, whereas that of pLDLR213mut/luc was unchanged (Fig. 4A). Mutation of the SRE decreased basal luciferase activity by 35% (p Ͻ 0.05) under the present culture conditions. Co-transfection of KLF13 suppressed wild-type pLDLR1076/luc by Ͼ85% (p Ͻ 0.01). If the latter value is normalized to 100%, then mutation at either Ϫ148 or Ϫ198 bp potentiated KLF13 repression by Ͼ60% (both p Ͻ 0.01 versus wild-type). Mutation of the central TCC triplet in the 5Ј-most distal Ϫ226/Ϫ202 sequence (pLDLR213mut/luc) did not alter inhibition by KLF13, whereas mutation of the SRE completely reversed repression by KLF13 and evoked 2-fold stimulation (p Ͻ 0.01 versus wild-type). These data suggest that KLF repression of the LDLR promoter may be antagonized by (a) other nuclear factor(s) acting via the two proximal TC-rich sites (such as Sp1 or Sp3) and (b) disruption of SREBP association with the centrally placed SRE. Because KLF13 has been previously shown to repress heterologous reporter promoters via a Sin3a-HDAC complex, using the same mutated reporter plasmids described above, we tested whether the HDAC inhibitor TSA had an effect on LDLR promoter activity (Fig. 4B). We observed that TSA increased the basal activity of wild-type pLDLR1076/luc by Ͼ2-fold compared with that quantitated in the absence of TSA. If the former response is defined as 100%, then mutations within Sp1-like sites at Ϫ148 and Ϫ198 (but not Ϫ213) diminished stimulation by TSA by Ͼ80% (p Ͻ 0.01) and Ͼ45% (p Ͻ 0.05), respectively. In contrast, compared with wild-type pLDLR1076/luc, mutation of the SRE enhanced the effect of TSA by 1.8-fold (p Ͻ 0.01). Compared with TSA alone, exposure to both KLF13 and TSA increased expression of pLDLR148mut and pLDLRSREmut by 2.1-and 1.9-fold, respectively. Together with the experiments described in the previous paragraph, the strong DNA context selectivity of promoter responses to an HDAC inhibitor described here raised the possibility that, at least functionally, there may exist a tripartite interaction among HDAC, KLF13, and SREBP.

KLF13 Repression of LDLR Promoter
To test this idea, we examined the effect of SREBP1 on KLF13-mediated repression of the LDLR promoter. Granulosa-luteal cells were cotransfected with varying concentrations of cytomegalovirus-driven N-terminal constitutively active SREBP-1a (0, 3, 10 ng of DNA), empty vector or a fixed amount of HisC-tagged human KLF13 (300 ng/well), and the reporter pLDLR1076/luc (600 ng) (Fig. 5A). Exogenous SREBP-1 alone increased luciferase activity by Ͼ7-fold, whereas KLF13 alone repressed reporter activity by Ͼ80% (both p Ͻ 0.01). Increasing amounts of SREBP-1 overcame suppression by KLF13. Western blot confirmed expression of KLF13 as shown in Fig. 5A, inset. To assess the specificity of transcriptional repression by KLF13, granulosa-luteal cells were transfected with CMV-driven Sp1 (100 ng of DNA), Sp3 (100 ng of DNA), or KLF13 (each 300 ng of DNA) and wild-type pLDLR1076/luc (600 ng of DNA). Sp1 and Sp3 each stimulated reporter activity by 2.0-fold. KLF13 reduced luciferase readout by 83 Ϯ 5% basally (p Ͻ 0.01) and by 46 Ϯ 4% when stimulated by Sp1 or Sp3 (p Ͻ 0.05). Concomitant exposure to TSA (10 ng/ml) elevated basal reporter expression by 2.5-fold (p Ͻ 0.01), completely reversed repression by KLF13, and augmented stimulation by Sp1 synergistically (by 4.1-fold) in the presence or absence of KLF13 (p Ͻ 0.05 versus TSA alone and p Ͻ 0.01 versus Sp1 plus TSA). TSA did not enhance the effect of Sp3 alone but doubled the response to combined Sp3 and KLF13 (Fig. 5B). Therefore, together these results support the hypothesis described above and permit a model in which the Sp1-SREBP complex must antagonize KLF13 to activate the LDLR promoter.
In addition to functional interactions between Sp1-SREBP and KLF13, we also investigated potential interactions between these proteins using gel retardation assays. In three assays, the addition of increasing amounts of recombinant SREBP-1a (1, 3, and 9 ng) decreased in vitro DNA binding of KLF13 to the Ϫ226/Ϫ202 oligo probe by 21, 25, and 83% (Fig. 6A). Thus, SREBP-1a suppresses the association of KLF13 with DNA in vitro. In contrast, SREBP-1 alone did not associate detectably with the Ϫ226/Ϫ202 oligo sequence (n ϭ 4 experiments). In vitro incubation of GST-KLF13 or GST with 35 S-SREBP-1 or HisC-SREBP-2 followed by GST pull-down and Western blotting disclosed a physical association between KLF13 and both SREBP-1 and SREBP-2 (Fig. 6B). N-terminal deletion of SREBP-1a abolished the in vitro interaction (Fig.  6C). Therefore, we next evaluated possible in vitro DNA-binding interactions between KLF13 and Sp1. Under EMSA conditions, the addition of GST-KLF13 increased the binding of intact Sp1 and the zinc finger motif of Sp1 to the 5Ј-most distal Ϫ226/Ϫ202 oligo DNA sequence (Fig.  7A). Specificity was affirmed by (a) supershift after pre-incubation with antibody to GST and (b) reduction of DNA-protein signal intensity after pre-incubation with antibody to KLF peptide. In confirmation of these results, increasing amounts of KLF13 enhanced the binding of fulllength Sp1 to the Ϫ226/Ϫ202 LDLR oligo (n ϭ 4 experiments) and to a consensus KLF13 (basic transcriptional element) DNA sequence (n ϭ 2 experiments). Conversely, pre-incubation with increasing amounts of Sp1 protein enhanced the association of a fixed concentration of KLF13 with the same oligo in EMSA (Fig. 7B). Therefore, collectively these results suggest that, in the absence of SREBP, Sp1 and KLF 13 may cooperate to bind the Sp1-like sequences within the LDLR promoter but that the repressor function of KLF13 must be dominant over the activation of Sp1. On the other hand, SREBP appears to be the protein necessary to displace KLF13 from the LDLR promoter. Accordingly, a physical interaction between these proteins may explain, at least in part, the functional interaction observed above.

DISCUSSION
During the last three decades, transcriptional regulation of the LDLR promoter has been demonstrated to be critical for cholesterol metabolism, such that hypercholesterolemic diseases arise from attendant defects. Currently, the most accepted model is that, in response to cholesterol deprivation, a functional complex formed by SREBP and Sp1 binds and activates the LDLR promoter, thereby increasing both the transcription of this gene and the amount of LDLR on the surface of responsive cells. In the current study, we extend this important paradigm 2-fold by showing first that there exists a default repressed state of the LDLR promoter, which must be overcome by the SREBP-Sp1 complex to activate its transcription. Second, we identify the Sin3a/HDACdependent repressor KLF13 as the protein that mediates such a function. This new paradigm must be taken into consideration not only when designing experiments but also in developing specific drugs that target these proteins. Therefore, the results presented here may have a high impact on accurately understanding altered cholesterol balance and posing pharmacological interventions.
Specifically, we have demonstrated that the product of the KLF13 gene binds each of three Sp1-like cis elements in the proximal (swine) LDLR promoter in a DNA context-selective manner in vitro, that it associates with the LDLR promoter in ovarian cells in an oxysterol-sensitive fashion in vivo, it strongly (Ͼ85%) represses basal transcription of an LDLR gene reporter   FEBRUARY 10, 2006 • VOLUME 281 • NUMBER 6 sequence, it antagonizes LDLR promoter stimulation by Sp1, Sp3, and SREBP-1 individually, and conversely, potentiates transcriptional activation induced by the Sp1-SREBP complex in the presence and absence of repressive amounts of exogenous sterol. Repression of the LDLR gene promoter by KLF13 is fully reversed by specific small interfering RNAs as well as by HDAC inhibitors. Given that ubiquitous Sp1 and tissue-specific SREBPs up-regulate the LDLR and other genes required for steroid-hormone or fatty acid biosynthesis (25,26,(37)(38)(39)(40), we suggest that KLF13 can inhibit and augment transcription of the LDLR promoter conditional to the availability of Sp1-SREBP.

KLF13 Repression of LDLR Promoter
TSA, a potent specific HDAC inhibitor (41), reversed repression of LDLR promoter activity by KLF13 and doubled the expression of endogenous LDLR gene transcripts in ovarian cells. The latter outcome is significant, because TSA up-regulated expression of Ͻ2% of 340 genes in human lymphoid cells (32). Deletional and mutational analyses of proximal LDLR promoter sequences revealed that repression by KLF13 and relief of repression by HDAC inhibitors require Sp1-like TC-rich sequences within the proximal Ϫ255/Ϫ139-bp region. The precise factors transducing promoter DNA context selectivity and the exact nature of the interaction between KLF13 and HDAC activity in this setting are not known. However, KLFs are able to enforce transcriptional silencing by recruiting HDAC and co-repressors, such as the mammalian homolog of Saccharomyces inhibitory sequence (yeast co-repressor protein) to certain promoters (42)(43)(44). In addition, Sp/KLF members synergize with nuclear co-activators, such as the acetyltransferases cyclic AMP-response element-binding protein and related p300 (9,27). Moreover, members of the Sp/KLF superfamily are acetylated and otherwise post-translationally modified, thereby predicting high functional specificity (45).
The proximal 5Ј upstream regions of the porcine and human LDLR genes differ in that the former harbors three distinct TC-rich sequences that flank a canonical SRE (29,46). Mutation of either of the two more proximal motifs in the pig gene reduced basal transcriptional activity by Ն50% but accentuated repression by exogenous KLF13, whether or not an HDAC inhibitor was present. The lack of reversal of inhibition by FIGURE 6. A, EMSA showing that in vitro incubation with increasing amounts (1, 3, and 9 ng) of SREBP-1a protein alone fails to form complexes with Ϫ226/Ϫ202 oligo DNA (left) but progressively decreases the binding of concomitantly added KLF13 protein (right). B, GST-KLF13 in vitro pull-down assay followed by immunoblotting of the SDS-PAGE-resolved proteins with anti-His antibodies to SREBP-1 and SREBP-2. C, deletion of the N terminus of SREBP-1 abolishes its in vitro association with KLF13. FIGURE 7. A, EMSA resulting from incubation of a Ϫ226/Ϫ202-bp LDLR oligo-DNA probe with GST-KLF13 or the zinc finger of Sp1 (Sp1-ZF). Pre-incubation was performed without or with antibody (Ab) to GST contained in GST-KLF13 (1) or to KLF13 contained in GST-KLF13 (2). The former Ab supershifted and the latter reduced the intensity of the 32 P-labeled protein-DNA complex. B, gel retardation assays demonstrating that increasing amounts of porcine KLF13 protein (in ng) enhance binding of recombinant human (rh) Sp1 (3 ng) to a consensus basic transcriptional element (BTE) oligo-DNA sequence (left) and to the TC-rich LDLR Ϫ226/Ϫ202 oligo (right). Both Sp1 and Sp3 enhance binding of a fixed (1 ng) amount of KLF13 to the Ϫ226/Ϫ202 oligo. Data are illustrative of similar outcomes in n ϭ 2 (BTE) and n ϭ 5 (LDLR) experiments.
TSA in these two contexts could indicate that these sites normally mediate gene activation, e.g. by Sp1, Sp3, or possibly KLF13, which can transduce either gene activation or repression (3,47). In contradistinction, mutation of the central TCC in the 5Ј-most upstream triple TCC repeat sequence in the LDLR gene did not modify inhibition by KLF13. Inactivation of the SRE lowered basal activity and doubled TSA-stimulated LDLR activity, both in the presence and absence of KLF13. These data raise the possibility that transcriptional inhibition requires interactions among KLF13, SREBP, and (unknown) non-acetylated co-repressor(s) and/or histones. Both Sp/KLF proteins and core nuclear histones are potential targets of acetylation and deacetylation (3,9,45,48). Although the precise nature of multimeric inhibition of the LDLR gene is not established, we show that the C-terminal zinc finger DNA-binding domain of porcine KLF13 is required for repression and that the N-terminal activational domain of SREBP is needed for the latter's in vitro association with KLF13 peptide.
Extractable nuclear and recombinant KLF13 preferentially bound the 5Ј-most TC-rich sequence (Ϫ244/Ϫ202) in the proximal LDLR promoter in gel retardation studies. Two other nuclear proteins, SREBP-1a and Sp1, modulated the cellular expression and action as well as DNA binding of KLF13 via distinct but complex mechanisms, inferred from the following. First, although SREBP-1 and KLF13 interacted physically in solution in GST pull-down assay, neither affected the other's in vitro binding to LDLR promoter DNA. Second, transfected SREBP-1a overcame KLF13 inhibition of LDLR reporter activity. Third, in contrast to SREBP, KLF13 enhanced Sp1 binding to the Ϫ226/Ϫ202 oligo and Sp1 augmented KLF13 association with the same sequence. Fourth, transfection of Sp1 or Sp3 attenuated KLF13 repression of the LDLR reporter and increased KLF13 protein expression. The last finding is consistent with Sp1 drive of other Sp/KLF gene promoters (49). This array of interactions confers a basis for elaborate feedback control of proximal LDLR promoter activity in vivo. Although further studies are needed to understand such dynamics, our data make clear that both KLF13 and Sp1-SREBP are involved.
In conclusion, the potent multifunctional transcriptional regulator KLF13 is expressed in untransformed ovarian cells, binds to three TCpredominant DNA sequences in the LDLR gene in vitro and to the proximal LDLR promoter in vivo, represses LDLR reporter activity via mechanisms that are sensitive to HDAC inhibition, and opposes LDLR promoter up-regulation by SREBP-1 or Sp1 individually. These observations suggest a novel model for LDLR gene repression, in which KLF13 mediates repression or activation under the control of SREBP-Sp1. This novel observation impacts valid framing of the basic biology, pathophysiology, and potential pharmacological manipulation of transcriptional regulation of the LDLR gene.